[Top] [All Lists]

Re: [PATCH] disable queue flag test in barrier check

To: Timothy Shimmin <tes@xxxxxxx>
Subject: Re: [PATCH] disable queue flag test in barrier check
From: Eric Sandeen <sandeen@xxxxxxxxxxx>
Date: Thu, 26 Jun 2008 08:25:11 -0500
Cc: xfs-oss <xfs@xxxxxxxxxxx>, LinuxRaid <linux-raid@xxxxxxxxxxxxxxx>, NeilBrown <neilb@xxxxxxx>, jeremy@xxxxxxxxx
In-reply-to: <48635284.3060001@xxxxxxx>
References: <486307EA.7080007@xxxxxxxxxxx> <48635284.3060001@xxxxxxx>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Thunderbird (Macintosh/20080421)
Timothy Shimmin wrote:

> Also from memory, I believe Neil checked this removal into the SLES10sp1 tree
> and some sgi boxes started having slow downs
> (looking at Dave's email below - we were not wanting to tell them
> to use nobarrier but needed it to work by default - I forget now).

But that's an admin issue.

The way it is now, for example a home user of md raid1 (me!) can't run
barriers even if they wanted to.

Until there is a way to know if a write cache is non-volatile the only
safe option is to enable barriers when possible.

> 6.
>> Date: Wed, 25 Jun 2008 08:57:24 +1000
>> From: Dave Chinner <david@xxxxxxxxxxxxx>
>> To: Eric Sandeen <sandeen@xxxxxxxxxxx>
>> Cc: LinuxRaid <linux-raid@xxxxxxxxxxxxxxx>, xfs-oss <xfs@xxxxxxxxxxx>
>> Subject: Re: md raid1 passes barriers, but xfs doesn't use them?
>> Yeah, the problem was that last time this check was removed was
>> that a bunch of existing hardware had barriers enabled on them when
>> not necessary (e.g. had NVRAM) and they went 5x slower on MD raid1
>> devices. Having to change the root drive config on a wide install
>> base was considered much more of support pain than leaving the
>> check there. I guess that was more of a distro upgrade issue than
>> a mainline problem, but that's the history. Hence I think we
>> should probably do whatever everyone else is doing here....
>> Cheers,
>> Dave.
> So I guess my question is whether there are cases where we are
> going to be in trouble again.
> Jeremy, do you see some problems?

FWIW, the problem *I* foresee is that some people are going to slow down
when using the defaults, yes, because barriers will start working again.
 But I don't see any other safe way around it.

Education would be in order, I suppose.  :)


> --Tim

<Prev in Thread] Current Thread [Next in Thread>